MACHINE LEARNING

ASSIGNMENT 2

Team

:R.Mallikarjun Reddy

Pyla Gagan

K.Himavanth Sai Sujan

Manikeshwar Reddy

Lalith Lavu

Issues in Machine Learning

•Data Quality and Quantity: Poor-quality or insufficient data leads to biased,

overfit, or underfit models, limiting generalization (e.g., non-diverse healthcare

datasets).

•Feature Selection: Irrelevant features harm model accuracy; automated systems

must refine predictive elements (e.g., color, texture in images).

•Hyperparameter Tuning: Optimizing hyperparameters is resource-intensive and

complex due to computational constraints.

•Overfitting and Underfitting: Overfitting occurs when a model captures noise and

inaccuracies from a large dataset, adversely affecting its performance.

Conversely, underfitting arises from a model being too simple for the data,

resulting in incomplete and inaccurate predictions.

TASK

Learning Stages in ML

Features: Models use features as independent variables for production

determination. Feature selection processes coupled with engineering

techniques provide substantial benefits to model predictive abilities.

Labels :The direction my model aims to determine consists of Labels

for prediction output. The model accuracy of supervised learning

depends strongly on the quality of provided labels. Achieving both

accurate and consistent labeling conditions fundamental for creating

effective models.

Hyperparameters: Hyperparameters represent adjustable

configuration settings which programmers adjust before starting the

learning process. Training becomes vulnerable to their influence

through controlled parameters because these settings regulate the

entire training timetable.

Validation: Model performance evaluation through validation occurs by

using the process to assess models against unknown data. Avoiding

overfitting requires appropriate validation approaches in decision-

making processes.

Real-Life Cases

Healthcare: According to Obermeyer et al. (2019) machine

learning encounters significant obstacles when used in

healthcare because training data proves to be biased. New

methods were developed to make data more diverse while

achieving fair models.

Natural Language Processing: Gururangan et al. (2018)

investigated how label noise affects sentiment analysis.

The study examined methods that could boost model

performance through better label consistency techniques.

References

Gururangan, S., Marasović, A., Swayamdipta, S.,

et al. (2018). Annotation artifacts in natural

language inference data. Proceedings of the

2018 Conference of the North American

Chapter of the Association for Computational

Linguistics: Human Language Technologies,

2018, 1-6.

Obermeyer, Z., Powers, B., Vogeli, C., &

Mullainathan, S. (2019). Dissecting racial bias

in an algorithm used to manage the health of

populations. Science, 366(6464), 447-453.

https://doi.org/10.1126/science.aax2342

TASK

•Is there a relationship between GPA and attendance?

•What about between SAT and attendance?

•Can you predict the attendance given the SAT score?

TASK

What probability should be used to make the

prediction?

The probability used to make a prediction is the

posterior probability for each class. The class with the

highest posterior probability is chosen as the prediction.

Naive Bayes uses the Bayes theorem to compute these

probabilities.

TASK

OUTPUT

TASK

Thank

You